Distributed photovoltaic modeling method and system based on Heiwen optimization extreme learning machine
By using the Black-winged Kite-based optimization of the Extreme Learning Machine (ELM) method, combined with Euclidean distance and grey relational analysis to screen similar days, and then using the improved Black-winged Kite algorithm to optimize the ELM model, the problems of insufficient accuracy and adaptability in distributed photovoltaic (PV) modeling are solved, and more efficient power prediction of distributed PV systems is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-20
- Publication Date
- 2026-04-14
AI Technical Summary
Existing technologies are insufficient for constructing accurate and efficient distributed photovoltaic models, cannot effectively cope with sudden weather changes, and lack versatility and adaptability.
A method based on the Black-winged Kite optimized Extreme Learning Machine was adopted. By combining Euclidean distance and grey relational analysis to calculate similarity, similar days were selected as input samples. The improved Black-winged Kite algorithm was then used to optimize the Extreme Learning Machine model for power prediction of distributed photovoltaic systems.
It improves the accuracy and adaptability of the model, enhances the model's convergence speed and generalization performance, ensures the quality of input data, and can better cope with the diversity and uncertainty of distributed photovoltaic systems.
Smart Images
Figure CN121863547A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of novel distributed photovoltaic modeling of power systems, specifically involving a distributed photovoltaic modeling method and system based on the Black-winged Kite Optimized Extreme Learning Machine. Background Technology
[0002] With the continuous increase in global energy demand, the development and application of new energy sources, and the vigorous development of renewable energy, can not only alleviate the energy crisis caused by the shortage of oil and gas resources, but is also an important strategy for the construction of a modern energy system. Solar power generation technology, as a mature and technologically advanced new energy power generation method in my country's clean energy development, has a very broad development prospect. However, due to the inherent randomness of solar energy, it has already posed a huge challenge to existing modern power systems with a high proportion of new energy sources.
[0003] Constructing accurate and efficient equivalent models of distributed photovoltaic (PV) systems to precisely control the dynamic characteristics of PV power plants in distribution networks is a crucial foundation for current distribution network planning, design, operation control, and system maintenance. However, distributed PV clusters contain numerous power plants with diverse dynamic characteristics, making it difficult for traditional modeling methods to resolve the contradiction between model accuracy and simulation speed. For example, while some models can fit historical data well, they often struggle to cope with sudden weather changes in real-world applications; furthermore, some models rely too heavily on datasets specific to particular scenarios, lacking sufficient generality and adaptability.
[0004] To address the aforementioned issues, there is an urgent need for a universal and adaptable distributed photovoltaic modeling method and system. Summary of the Invention
[0005] To address the shortcomings of existing technologies, this invention provides a distributed photovoltaic modeling method and system based on the Black-winged Kite Optimized Extreme Learning Machine.
[0006] The present invention specifically adopts the following technical solution.
[0007] In a first aspect, the present invention discloses a distributed photovoltaic modeling method based on the Black-winged Kite Optimized Extreme Learning Machine, comprising the following steps: S1. Collect environmental data from each photovoltaic power station and simultaneously acquire power data corresponding to the time of environmental data collection. The environmental data includes irradiance and ambient temperature. S2. Calculate the similarity coefficient between the target day's environmental data and historical day's environmental data. Based on the similarity coefficient, select similar days as input samples, and the number of similar days is adaptively adjusted based on historical power fluctuations. S3. Input the power data in the input sample into the grey theory model to obtain the fitted value, and correct the fitting error based on the Markov algorithm to obtain the corrected fitted power sequence. S4. Input the corrected fitted power sequence and the corresponding environmental data into the Extreme Learning Machine model to obtain the power prediction results of the distributed photovoltaic system. The Extreme Learning Machine model is optimized using the improved Blackwing Kite algorithm.
[0008] More preferably, The calculation of the similarity coefficient between the target day's environmental data and historical day's environmental data is specifically as follows: Calculate the Euclidean distance similarity and grey relational coefficient between the target day's environmental data and historical day's environmental data; weight the similarity coefficients. Multiply by the Euclidean distance similarity, then subtract the similarity weight coefficient from 1. The result is multiplied by the grey relational coefficient, and the sum of the two multiplications is used as the similarity coefficient.
[0009] More preferably, The similarity weighting coefficient increases as the average forecast value of the target day's irradiance increases.
[0010] More preferably, The process of selecting similar days as input samples based on similarity coefficients is as follows: Sort all historical days in descending order based on similarity coefficient, and select the top... N A number of historical dates were used as similar dates and as input samples.
[0011] More preferably, The number of similar days is adaptively adjusted based on historical power fluctuations, and is determined in the following manner:
[0012] In the formula, N The number of similar days. , These are the minimum number of similar days and the maximum number of similar days, respectively. It is the sigmoid function; V Historical power fluctuations; This is the slope parameter, used to control the transition speed; for V The midpoint threshold; where, and Calibration is trained using historical data.
[0013] More preferably, The improved Black-winged Kite algorithm specifically includes: The position of each generation of black-winged kites is updated by perturbation using a mathematical model that simulates rabbits foraging around obstacles in the artificial rabbit optimization algorithm; and the search step size is adjusted by using the system fluctuation coefficient.
[0014] More preferably, The system fluctuation coefficient is the average of the standard deviations of power, temperature, and irradiance within a preset sliding window length.
[0015] Secondly, the present invention discloses a distributed photovoltaic modeling system based on the aforementioned method, including a data acquisition module, an input sample extraction module, a fitted power sequence calculation module, and a distributed photovoltaic system power prediction module; The data acquisition module collects environmental data from each photovoltaic power station and simultaneously acquires the power data corresponding to the time of environmental data acquisition. The environmental data includes irradiance and ambient temperature. The input sample extraction module calculates the similarity coefficient between the target day's environmental data and historical day's environmental data, selects similar days as input samples based on the similarity coefficient, and adaptively adjusts the number of similar days based on historical power fluctuations. The power sequence fitting module inputs the power data from the input sample into the grey theory model to obtain the fitting value, and corrects the fitting error based on the Markov algorithm to obtain the corrected power sequence. The distributed photovoltaic system power prediction module inputs the corrected fitted power sequence and the corresponding environmental data into the extreme learning machine model to obtain the distributed photovoltaic system power prediction result. The extreme learning machine model is optimized using the improved Blackwing Kite algorithm.
[0016] Thirdly, the present invention provides a terminal, including a processor and a storage medium; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method according to any one of the first aspects of the present invention.
[0017] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in any of the first aspects of the present invention.
[0018] The beneficial effects of this invention are compared with those of the prior art: This invention presents a distributed photovoltaic modeling method and system based on the Black-winged Kite Optimized Extreme Learning Machine. By combining two similar-day clustering methods—Euclidean distance and grey relational analysis—and using irradiance and temperature as variables, it provides more complementary data input for subsequent models. When calculating the similarity between the target day and historical days, this invention categorizes weather scenarios into three types based on real-time irradiance and dynamically adjusts the weights of the corresponding coefficients for the Euclidean distance and grey relational analysis clustering methods. This ensures that the similarity calculation accurately matches the core logic of power fluctuations under different irradiance levels. Simultaneously, by adopting an adaptive adjustment of the number of similar days based on historical power fluctuations, it accurately matches the characteristics of the target day, solving the problems of rigid sample selection and poor adaptability in existing methods. By selecting the most representative similar-day data through the above methods, the quality of the input data is ensured. This invention proposes a distributed photovoltaic modeling method based on the Black-winged Kite Optimized Extreme Learning Machine. The method utilizes the Black-winged Kite optimization algorithm to optimize the parameters of the Extreme Learning Machine, thereby improving the convergence speed and generalization performance of the model and obtaining more accurate results. Attached Figure Description
[0019] Figure 1 This is a flowchart of the distributed photovoltaic modeling method based on the Black-winged Kite Optimized Extreme Learning Machine of the present invention. Detailed Implementation
[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of this invention. The embodiments described in this application are merely some embodiments of this invention, and not all embodiments. Based on the spirit of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of this invention.
[0021] Example 1: like Figure 1 As shown, this invention discloses a distributed photovoltaic modeling method based on the Black-winged Kite Optimized Extreme Learning Machine, comprising the following steps: S1. Collect environmental data from each photovoltaic power station and simultaneously acquire power data corresponding to the time of environmental data collection. The environmental data includes irradiance and ambient temperature. S2. Calculate the similarity coefficient between the target day's environmental data and historical day's environmental data. Based on the similarity coefficient, select similar days as input samples, and the number of similar days is adaptively adjusted based on historical power fluctuations. Considering that the output power of individual photovoltaic (PV) power plants in a distributed photovoltaic (PV) system may vary due to environmental factors, it is first necessary to perform personalized filtering of historical data for each PV power plant. When selecting similar days, irradiance and ambient temperature should be considered to more accurately identify historical dates that match the conditions. The specific steps are as follows: A1. Obtain numerical weather forecast information on irradiance and temperature for the target day and perform standardization processing. Let the standardized irradiance vector for the target day be... Temperature vector is The expression is as follows:
[0022]
[0023] in, n This represents the total number of sampling points within a single day.
[0024] A2. Let the standardized number be... m The irradiance vector for each historical day is Temperature vector is The expression is as follows:
[0025]
[0026] in, n This represents the total number of sampling points within a single day.
[0027] A3. Calculate the Euclidean distance between the target date and historical dates, and perform similarity conversion, specifically as follows:
[0028] In the formula, For the target date and the m Euclidean distance over a historical day; For the target day k Irradiance at each sampling point For the first m The first historical day k Irradiance at each sampling point For the target day k Temperature at each sampling point For the first m The first historical day k Temperature at each sampling point n This represents the total number of sampling points within a single day. Furthermore, a similarity transformation is performed on the Euclidean distance between the target date and the historical date, specifically as follows:
[0029] In the formula, The similarity between the target day's environmental data and historical day's environmental data is calculated using the Euclidean distance.
[0030] A4. Calculate the grey relational coefficients of irradiance and temperature for the target day and historical days:
[0031]
[0032] In the formula, For the target day and the m Grey relational coefficient of historical daily irradiance; For the target day and the m Grey relational coefficient of historical daily temperature; For the target day k Irradiance at each sampling point For the first m The first historical day k Irradiance at each sampling point For the target day k Temperature at each sampling point For the first m The first historical day k Temperature at each sampling point The minimum absolute difference between the irradiance values of all historical days and the target day. The maximum absolute difference between the irradiance values of all historical days and the target day. The minimum absolute difference between the temperature differences of all historical days and the target day. The maximum absolute difference between the temperature values of all historical days and the target day. , The resolving factor is used to adjust the sensitivity of the correlation coefficient to differences.
[0033] The average of the grey relational coefficients for irradiance and temperature is taken as the final grey relational coefficient, specifically:
[0034] In the formula, The final grey relational coefficient is the environmental data for the target day and the environmental data for historical days.
[0035] A5. Calculate the similarity coefficient between the target date and historical dates, specifically as follows:
[0036] In the formula, The similarity coefficient between the target date and historical dates; This refers to the similarity weighting coefficient; Specifically, the similarity weighting coefficient is dynamically adjusted based on real-time weather scenarios, increasing as the average daily irradiance forecast for the target day increases. Preferably, it can be determined in the following manner: When the forecast mean daily irradiance G > G1, the daily average irradiance is high, and the irradiance value dominates the similarity. The similarity weighting coefficient focuses on the numerical difference. The value is 0.7; When the forecast mean of the target day's irradiance is G≤G1 and G>G2, the numerical fluctuation and trend are in equilibrium. The value is 0.5; When the forecast mean daily irradiance G ≤ G2, temperature has a significant impact on power, and the focus is on trend consistency. The value is 0.3.
[0037] A6. Historical days are filtered based on similarity coefficients to obtain similar days, which are then used as input samples; the number of similar days is adaptively adjusted based on historical power fluctuations. Specifically, according to the similarity coefficient Sort all historical dates in descending order and select the previous ones. N A historical date is used as a similar date; Furthermore, the number of similar days N Determine as follows:
[0038] In the formula, , These are the minimum number of similar days and the maximum number of similar days, respectively. For the sigmoid function, ensure the output is between 0 and 1, and implement... N from arrive Smooth transition; V Historical power fluctuations are used to quantify the stability of recent power. The slope parameter controls the transition speed; the larger the value, the faster the transition. V Caused by tiny changes N The more drastic the change; for V The midpoint threshold; where, and Calibration is trained using historical data; Preferably, historical power fluctuations V Target date h The power standard deviation of the day.
[0039] S3. Input the power data in the input sample into the grey theory model to obtain the fitted value, and correct the fitting error based on the Markov algorithm to obtain the corrected fitted power sequence. Due to the complexity and diversity of distributed photovoltaic (PV) systems, the data specificity among different power plants must be considered when selecting a model. A model constructed using grey theory combined with Markov chains can effectively handle the output power fluctuations of each power plant. Based on this, the parameter settings of the grey-Markov model can be further adjusted according to the specific conditions of each power plant to improve accuracy. This step focuses on optimizing the model's adaptability to changes in the output power of different distributed PV power plants.
[0040] The Markov algorithm is added to the grey theory model to correct errors and further optimize it. The specific process is as follows: B1, the first in the input sample m The true value of similar daily power data; Specifically, the first input sample m A similar daytime first k The true power value of each sampling point is ,in, n This represents the total number of sampling points within a single day. M This represents the total number of similar days in the input sample.
[0041] B2. Obtain the first [model] generated by the grey theory model. m Fitted values of power data from several similar days; Specifically, the grey theory model generates the first m A similar daytime first k The fitted value of the power data at each sampling point is ,in, n This represents the total number of sampling points within a single day. M This represents the total number of similar days in the input sample.
[0042] B3. Calculate the error between the actual value of similar days and the fitted value of the grey theory model. for:
[0043] B4. Because the power fitting values for similar days and the fitting values for the target day are both sequences generated by the grey theory model, and the errors between the two and the true values have the same arrangement, the error sequence between the fitting values and the true values of similar days is set as the input data of the Markov chain; the errors of all similar days are... After statistics, it was divided into C The nth non-overlapping state intervals, the th i Interval expression for each state for:
[0044]
[0045]
[0046] In the formula, , These are the upper and lower limits of the error in the transition interval. , This is the transformation constant.
[0047] After obtaining the state interval, the state transition probability matrix is calculated using the Markov chain. The corrected fitted values for similar days are obtained as follows:
[0048] In the formula, , Based on the state transition probability matrix Iterative correction.
[0049] S4. Input the corrected fitted power sequence and the corresponding environmental data into the Extreme Learning Machine model to obtain the power prediction results of the distributed photovoltaic system. The Extreme Learning Machine model is optimized using the improved Blackwing Kite algorithm. To address the characteristics of distributed photovoltaic (PV) systems, a multi-strategy improved Black-winged Kite BKA optimization algorithm is proposed to optimize key parameters in the Extreme Learning Machine (ELM) model. By adjusting the hidden layer weights and biases, the model not only achieves faster convergence but also enhances its generalization ability, thereby better handling the diversity and uncertainty of output power from various power plants in distributed PV systems. The improved BKA algorithm is incorporated into ELM, and the specific process is as follows: C1, the Black-winged Kite algorithm initializes the population by randomly generating a set of initial solutions in space:
[0050] In the formula, This is the initial solution obtained after initialization. A random value between 0 and 1 and These represent the upper and lower boundaries of the search range, respectively.
[0051] The black-winged kite adjusts the angle of its wings and tail according to wind speed during flight, then swiftly dives to attack; the mathematical expression is:
[0052]
[0053] in, , The first t Second and third tThe position of the Black-winged Kite individual at +1 iteration; A random number ranging from 0 to 1; The value is a preset probability constant, and in this embodiment, it is preferably set to 0.9; The maximum number of iterations, This represents the current iteration number. n This is the step size adjustment factor.
[0054] C2. Initializing the population using a mirror back learning strategy can quickly search for a richer population, thereby improving the efficiency of algorithm iteration and optimization; the specific formula is as follows:
[0055]
[0056]
[0057] in, To initialize the population using a mirror back learning strategy, the first... i The initial solution for each individual; , These are the upper and lower bounds of the current dimension, respectively; As a control parameter, its value is determined by... , and Decision, among which , and All are random numbers uniformly distributed between 0 and 1; For the population dimension.
[0058] C3. To improve the overall optimization performance and convergence speed of the BKA algorithm, during the flight of the black-winged kite, the mathematical model simulating rabbits detouring to forage in the artificial rabbit optimization algorithm is used to perturb and update the position of each generation of black-winged kites. At the same time, a system fluctuation coefficient is proposed to measure the degree of fluctuation of the data under the current photovoltaic scenario. When the fluctuation is small, that is, the current regularity is strong, a local fine search is performed. When the fluctuation is large, that is, the uncertainty is high, a global broad search is performed. Specifically, when At that time, the black-winged kite's detour foraging strategy position update formula is as follows:
[0059]
[0060]
[0061]
[0062] in, The remaining individual black-winged kites represent the positions of the remaining solution vectors. To perform rounding operations on numerical values, For random numbers that follow a standard distribution, and A random value between 0 and 1. Let be the perturbation factor, and e be the natural constant. This represents the current iteration number. The maximum number of iterations, This means randomly selecting an integer between 1 and D; For system fluctuation coefficient, , , These are the standard deviations of power, temperature, and irradiance, respectively, calculated from historical power, temperature, and irradiance data within a sliding window (such as the most recent 10 samples). The improved method for updating the individual positions of black-winged kites proceeds in both forward and reverse directions, adjusting the step size based on the number of iterations. Initially, a larger step size is chosen for flight, while a smaller step size is used later for precise position updates. Simultaneously, the fluctuation level of data in a photovoltaic scenario is incorporated to achieve adaptive amplification / attenuation of disturbance intensity. This effectively improves the global optimization capability and convergence speed of the BKA algorithm, preventing it from getting trapped in local optima.
[0063] C4. A cross-sectional strategy is introduced to optimize the BKA algorithm. Horizontal cross-sectional ... Its value dynamically decreases as the number of iterations increases, balancing exploration and development to achieve large-scale parameter exploration in the early stages and fine-tuning in the later stages; the specific formula is as follows:
[0064]
[0065] in, The initial variation amplitude, t This represents the current iteration number. T This represents the total number of iterations. The attenuation coefficient is used to control... r The rate of decay; The current optimal individual; For the first i The first individual j dimensional variables, , The first i +1, No. i -1 individual's first j Dimensional variables.
[0066] Example 2: This invention also claims protection for a distributed photovoltaic modeling system based on the aforementioned method, comprising a data acquisition module, an input sample extraction module, a fitted power sequence calculation module, and a distributed photovoltaic system power prediction module; The data acquisition module collects environmental data from each photovoltaic power station and simultaneously acquires the power data corresponding to the time of environmental data acquisition. The environmental data includes irradiance and ambient temperature. The input sample extraction module calculates the similarity coefficient between the target day's environmental data and historical day's environmental data, selects similar days as input samples based on the similarity coefficient, and adaptively adjusts the number of similar days based on historical power fluctuations. The power sequence fitting module inputs the power data from the input sample into the grey theory model to obtain the fitting value, and corrects the fitting error based on the Markov algorithm to obtain the corrected power sequence. The distributed photovoltaic system power prediction module inputs the corrected fitted power sequence and the corresponding environmental data into the extreme learning machine model to obtain the distributed photovoltaic system power prediction result. The extreme learning machine model is optimized using the improved Blackwing Kite algorithm.
[0067] Example 3: An embodiment of the present invention provides a terminal, including a processor and a storage medium; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method according to any one of Embodiment 1.
[0068] Example 4: The present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in any one of the embodiments.
[0069] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions of the present invention still include content that can be modified or equivalently replaced in the specific implementation of the present invention. Any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.
Claims
1. A distributed photovoltaic modeling method based on the Black-winged Kite Optimized Extreme Learning Machine, characterized in that, Includes the following steps: S1. Collect environmental data from each photovoltaic power station and simultaneously acquire power data corresponding to the time of environmental data collection. The environmental data includes irradiance and ambient temperature. S2. Calculate the similarity coefficient between the target day's environmental data and historical day's environmental data. Based on the similarity coefficient, select similar days as input samples, and the number of similar days is adaptively adjusted based on historical power fluctuations. S3. Input the power data in the input sample into the grey theory model to obtain the fitted value, and correct the fitting error based on the Markov algorithm to obtain the corrected fitted power sequence. S4. Input the corrected fitted power sequence and the corresponding environmental data into the Extreme Learning Machine model to obtain the power prediction results of the distributed photovoltaic system. The Extreme Learning Machine model is optimized using the improved Blackwing Kite algorithm.
2. The distributed photovoltaic modeling method based on the Black-winged Kite Optimized Extreme Learning Machine according to claim 1, characterized in that, The calculation of the similarity coefficient between the target day's environmental data and historical day's environmental data is specifically as follows: Calculate the Euclidean distance similarity and grey relational coefficient between the target day's environmental data and historical day's environmental data; weight the similarity coefficients. Multiply by the Euclidean distance similarity, then subtract the similarity weight coefficient from 1. The result is multiplied by the grey relational coefficient, and the sum of the two multiplications is used as the similarity coefficient.
3. The distributed photovoltaic modeling method based on the Black-winged Kite Optimized Extreme Learning Machine according to claim 2, characterized in that, The similarity weighting coefficient increases as the average forecast value of the target day's irradiance increases.
4. The distributed photovoltaic modeling method based on the Black-winged Kite Optimized Extreme Learning Machine according to claim 2, characterized in that, The process of selecting similar days as input samples based on similarity coefficients is as follows: Sort all historical days in descending order based on similarity coefficient, and select the top... N A number of historical dates were used as similar dates and as input samples.
5. The distributed photovoltaic modeling method based on the Black-winged Kite Optimized Extreme Learning Machine according to claim 4, characterized in that, The number of similar days is adaptively adjusted based on historical power fluctuations, and is determined in the following manner: In the formula, N The number of similar days. , These are the minimum number of similar days and the maximum number of similar days, respectively. It is the sigmoid function; V Historical power fluctuations; This is the slope parameter, used to control the transition speed; for V The midpoint threshold; where, and Calibration is trained using historical data.
6. The distributed photovoltaic modeling method based on the Black-winged Kite Optimized Extreme Learning Machine according to claim 5, characterized in that, The improved Black-winged Kite algorithm specifically includes: The position of each generation of black-winged kites is updated by perturbation using a mathematical model that simulates rabbits foraging around obstacles in the artificial rabbit optimization algorithm; and the search step size is adjusted by using the system fluctuation coefficient.
7. The distributed photovoltaic modeling method based on the Black-winged Kite Optimized Extreme Learning Machine according to claim 6, characterized in that, The system fluctuation coefficient is the average of the standard deviations of power, temperature, and irradiance within a preset sliding window length.
8. A distributed photovoltaic modeling system based on the method of any one of claims 1-7, comprising a data acquisition module, an input sample extraction module, a fitted power sequence calculation module, and a distributed photovoltaic system power prediction module, characterized in that: The data acquisition module collects environmental data from each photovoltaic power station and simultaneously acquires the power data corresponding to the time of environmental data acquisition. The environmental data includes irradiance and ambient temperature. The input sample extraction module calculates the similarity coefficient between the target day's environmental data and historical day's environmental data, selects similar days as input samples based on the similarity coefficient, and adaptively adjusts the number of similar days based on historical power fluctuations. The power sequence fitting module inputs the power data from the input sample into the grey theory model to obtain the fitting value, and corrects the fitting error based on the Markov algorithm to obtain the corrected power sequence. The distributed photovoltaic system power prediction module inputs the corrected fitted power sequence and the corresponding environmental data into the extreme learning machine model to obtain the distributed photovoltaic system power prediction result. The extreme learning machine model is optimized using the improved Blackwing Kite algorithm.
9. A terminal, comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method according to any one of claims 1-7.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method according to any one of claims 1-7.